The price drop is probably more interesting than the benchmark improvement.
At these prices, you can start throwing Flash at a lot of small, repetitive tasks where you wouldn't even consider using a bigger model before. It feels like the interesting shift is not “Flash replaces Pro”, but “there are now a lot more things worth automating.”
Yes, my favorite codebase to work in is one full of accurate and long comments that detail that making a function call it never called would cause problems because of the details of a function that gets called 3 layers deeper.
I think the bigger issue is that AI is attacking the business model, not necessarily the product.
Tailwind can still be massively popular, but if an AI agent can look at the docs/examples and generate 80% of what you'd otherwise pay for, that popularity doesn't translate into customers anymore.
Feels like we'll see this happen to a lot of "we already figured this out for you" developer products.
> Feels like we'll see this happen to a lot of "we already figured this out for you" developer products.
It has for me at least.
I had a solo business for 10 years selling tech video courses (learning Flask, Docker, etc.). It was enough to pay the bills, but for the last few years it has nose dived. For 2026 YTD it has generated $20 / month in revenue but it costs almost triple that to host the courses so I've been losing hundreds of dollars a year, for years. All while watching LLMs cannibalize ~600 free in-depth blog posts and YouTube videos only to re-sell that information to anyone willing to pay.
It completely flipped my life upside down to where for 20 years I was able to sustain myself doing contract work + courses to having to take full time work. All of this started when AI became popular.
It's all related to a massive loss of traffic to my site which was the entry point to both contract work and the courses. Both sides of the business have been crushed.
Often time I just found myself unconsciously take out my phone and start scrolling for no reason. Was not a conscious decision just muscle memory. Stop scrolling isn't really an option for me because a lot of info are genuinely useful, but I do fall into rabbithole of some of the most random stuff. Feel like that's how most people start doom scrolling
Other than the pricing, for something built for simplicity, i always found their config method bit weird. Maybe it's just me b/c I first learned the aws stack, and got tired of managing EKS
Yeah I don't see a reason to use k8s if not on cloud, and k8s on cloud is even more complex. Compose is still the right size for one box, that one command does all was really nice to have
Agentic product is nice to have if im new to it, but for simple products I'm familiar with, having to put my thought into words is way more annoying than typing in a couple numbers in a UI. I always thought "agent with tools" is just a new version of UI which is a info delivery method.
But on a second thought, most people in a company is essentially "agent with tools", so maybe it's already pretty amazing
Yeah maybe I'm conflating an "agent with tools" with a chat feature. Just because I'm using an agent with tools doesn't mean I need to chat with it, I could just click a button and have it do things, in the same way that I click a button and a driver with tools goes and gets my food.
I'm debating the tradeoff for secret management in my app as well. Storing it is easy you just need encryption and it's mostly good. But delivering it is tricky. Delivery via env is simple for sure but can get leaked. The other route would be a job scoped signature, but this doesnt stop the job from printing the secret out, it only shrinks the blast radius.
But if you delete the secret after the job is done or deployment is up, it's pretty much the same result
This is similar to my experience when cursor first came out. But now it's even trickier because LLM actually writes working code that pass the test. The issue is when i dig in deeper, there's often small security or design issues that will cause problem down the road when the app gets more traffic.
These issues are usually in really tedious places that I didn't bother to describe what to do to the agent
> design issues that will cause problem down the road when the app gets more traffic.
Right but that is someone else's problem. Or if that's something you will be call upon to answer, "that's just LLM did" is a fine answer in most companies where AI use has been mandated.
At these prices, you can start throwing Flash at a lot of small, repetitive tasks where you wouldn't even consider using a bigger model before. It feels like the interesting shift is not “Flash replaces Pro”, but “there are now a lot more things worth automating.”